Thunderbolt 5 Drive Testing

[Hugh Barstead]

Using a Mac Studio M3 Ultra (standard config with 28 core CPU, 96 GB RAM and 1 TB HD), additional storage was required. The solution was to employ an external M.2 drive in a Thunderbolt 5 enclosure for maximum speed.

There are few choices in external drives with TB5 connectivity. A couple of complete drives and a couple of enclosures were all that were available for immediate order.

Of those, an enclosure from Acasis was the only one incorporating a cooling fan so it was procured, model number TB501Pro, advertising 80 Gb/s speed.

The requirements for the M.2 NVME drive are specific. It must be a 4 lane stick employing PCIe Gen 4. The enclosure does not support Gen 5.

On hand was a WD_Black SN850X NVMe™ SSD, 2TB. It fit all the requirements and seemed fast enough so it was installed in the Acasis.

The drive was plugged into the Mac using the Acasis supplied cable and formatted with the Apple disk utility in APFS to match the internal drive. Approximately a half TB of miscellaneous data files were then transferred to the drive prior to testing to better simulate real world conditions.

Blackmagic’s Disk Speed Test was used to check the throughput. This software was selected due to familiarity.

The results are noteworthy. To begin with, both the internal and external drives wrote faster than they read. No fluke, repeated testing and reinstalling the testing software yielded exactly the same results.

Those results were fast. First, the internal 1 TB drive:




Internal Drive


Then the external 2 TB Thunderbolt 5 drive:




External Drive



Note: When testing, no two results were identical. The widest variation was on the order of 8%. The results shown are “typical” in that they are close to the mean.

To complete the testing a 138.1 GB .NEV file was transferred from the external to the internal drive using a Finder copy. Timing the transfer was considered accurate to within 0.1 second.

The transfer took 22.93 seconds which results in a 6.023 GB/S transfer speed. This is actually slightly faster than the results from the Blackmagic testing (taking the lower read speed into account), likely due to the single monolithic file transfer.

Transferring the same file from the internal to the external drive took slightly longer, resulting in a transfer speed of 5.4 GB/S. Speculation was that this was due to the slower read speed of the internal drive, however this is just an uninformed guess.

While the TB5 enclosures are still quite expensive (the Acasis cost ~ $225 U.S., plus the cost of the drive) it is still far less costly than Apple’s drive pricing.

With the TB5 drive closely matching the speed of the internal drive it makes sense to go this route rather than paying Apple. NB: The larger drives provided by Apple may or may not be significantly faster than the 1 TB drive contained in the test Mac Studio. Any info on this would be appreciated.

In comparison a PC equipped with a USB-C 3.2 Gen 2x2 card and an appropriate drive and cable attached delivered 1.8 GB/S under similar test conditions. This amounts to 72% of the theoretical 20 Gb/S speed available.

TB5 appears to be three times faster while using only 60% of the theoretical throughput. Perhaps efficiency gains in the future will provide even more speed.

Certainly there is much to criticize in the testing methods employed. The write speeds that are faster than the read speeds is unusual. However experience in real world usage shows that the tested speeds are realized in practice. Perceived speeds of the internal and external drives are essentially identical.

Input from anyone with TB5 drive experience is welcome.

 

[RLight]

Interesting, and not surprising.

Question? If you're not opposed to the command line and home-brew, could you provide the latency difference between the Thunderbolt 5 connected drive and your internal drive? You'd need to install and run FIO though to produce this. I'm curious the differences in bus-attachment latency between the two. Although I suspect the throughput of the ThunderBolt 5 with a modern NVME should exceed the internal storage of the Mac, the latency might not due to the closer proximity on the bus of the internal NVME. But, it could be that the external NVME attached via TB5, exceeds both.

 

[wh1tby]

I have an M4 Pro Mac Mini (64 GB with 2 TB SSD) which has 3 TB5 ports. Unfortunately i bought it before the release of the M4 based Studio series which I would have preferred. I had been using a Trebleet TB5 enclosure with a Samsung 990 Pro 4TB SSD successfully and achieved similar speeds to yours. The only issue I had was that this enclosure has a permanently running fan which made me have concerns about long term reliability as it runs 24/7 (even when the machine was in sleep mode).

I tried the Acasis TB501 Pro ( which has a temperature controlled fan) and the new OWC Express 1M2 80G fan less enclosure and eventually settled on the OWC which runs very cool and has no fans. Speeds are almost indistinguishable from yours. The Acasis enclosure seems less robust, has less heatsink mass than either the Trebleet or OWC and costs more than either of the others.

As you point out the total cost is far less than Apple charge for their storage and achieves identical speeds.

 

[MarshallG]

Black magic is a wonderful company and their app is very cool looking, but it’s no longer considered the gold standard for drive benchmarking. Instead, try the app AmorphousDiskMark. Just as easy to use.



As with BlackMagic, there are a few configuration settings, mainly the block size used in the test (I think most people use the largest). But this should give you more useful and accurate results. I think you’ve got useful results here. TB5 still new, so we can expect prices to fall over the next few years.

 

[RLight]

Black magic is a wonderful company and their app is very cool looking, but it’s no longer considered the gold standard for drive benchmarking. Instead, try the app AmorphousDiskMark. Just as easy to use.

As with BlackMagic, there are a few configuration settings, mainly the block size used in the test (I think most people use the largest). But this should give you more useful and accurate results. I think you’ve got useful results here. TB5 still new, so we can expect prices to fall over the next few years.

No disrespect to CrystalMark clones, but FIO, is the gold standard. There’s a reason TomsHardware and data center support uses it… Install with Brew. It can reveal idiosyncrasies CrystalMark misses. Now Crystal Mark is better than Black Magic.

 

[wh1tby]

Black magic is a wonderful company and their app is very cool looking, but it’s no longer considered the gold standard for drive benchmarking. Instead, try the app AmorphousDiskMark. Just as easy to use.

As with BlackMagic, there are a few configuration settings, mainly the block size used in the test (I think most people use the largest). But this should give you more useful and accurate results. I think you’ve got useful results here. TB5 still new, so we can expect prices to fall over the next few years.

I use Black Magic, AJA System Test and AmorphousDiskMark to give me an overview of performance. e.g. is the interface working as expected, cache sizes etc. In the end I use a test with Lightroom of a specific version ingesting a specific set of 200 Sony A7RV raw files to see how the overall system performs and a Carbon Copy Cloner task for copying a large set of files (about 1 TB).

By using the same version of LRc and CCC and the same files, a comparison can be made that indicates any issues with caching etc. and allows me to discount the differences in application and files to give me a real world indication of how the new drive will perform. When a new version of either LR or CCC comes out, I simply rerun the test on all drives to get a new baseline. But in the end, disk transfer speed improvements are only useful if the applications we use can make use of them.

TB5 is definitely a lot faster than TB 4 (in my testing it results in about 85% improvement using my application testing) which makes the investment in TB5 compliant interfaces worth the money, even though TB4 drive enclosures are about 45% the cost of TB5 enclosures.

 

[MarshallG]

Black magic is a wonderful company and their app is very cool looking, but it’s no longer considered the gold standard for drive benchmarking. Instead, try the app AmorphousDiskMark. Just as easy to use.

As with BlackMagic, there are a few configuration settings, mainly the block size used in the test (I think most people use the largest). But this should give you more useful and accurate results. I think you’ve got useful results here. TB5 still new, so we can expect prices to fall over the next few years.

I use Black Magic, AJA System Test and AmorphousDiskMark to give me an overview of performance. e.g. is the interface working as expected, cache sizes etc. In the end I use a test with Lightroom of a specific version ingesting a specific set of 200 Sony A7RV raw files to see how the overall system performs and a Carbon Copy Cloner task for copying a large set of files (about 1 TB).

By using the same version of LRc and CCC and the same files, a comparison can be made that indicates any issues with caching etc. and allows me to discount the differences in application and files to give me a real world indication of how the new drive will perform. When a new version of either LR or CCC comes out, I simply rerun the test on all drives to get a new baseline. But in the end, disk transfer speed improvements are only useful if the applications we use can make use of them.

TB5 is definitely a lot faster than TB 4 (in my testing it results in about 85% improvement using my application testing) which makes the investment in TB5 compliant interfaces worth the money, even though TB4 drive enclosures are about 45% the cost of TB5 enclosures.

Thank you! That’s great to know.
85% is probably the maximum we can expect, wouldn’t you say? Or perhaps a full 100%. I’m surprised there are solutions delivering that well so soon after introduction. I’m sure prices of enclosures will come down in time; unless tariffs remain high.

 

[CAcreeks]

Of those, an enclosure from Acasis was the only one incorporating a cooling fan so it was procured, model number TB501Pro, advertising 80 Gb/s speed.

Is the fan optional? Looks like there is a plug for it, marked with a somewhat fan-like icon, but no pictures show anything attached there. Never mind. It's a button switch. Quote:

Regarding the fan on the TB501 Pro:
Manual Control: This black button is used to adjust the fan. You need to press and hold the button for one second to turn the fan on. Press it again to turn off the fan.
Smart Temperature Control: The TB501 Pro features an intelligent thermal management system that activates the fan automatically when the internal temperature reaches a certain threshold, even without pressing the button. This dual-control system (manual and automatic) ensures optimal cooling while maintaining efficient operation.

 

[LarryRC]

There are other TB 5 enclosures with a fan such as the UGREEN 80Gbps M.2 NVMe SSD Enclosure with Cooling Fan, Thunderbolt 5 enclosure.

This is the one I have been using for about 3 months and it, so far, has been rock solid. I use a WD_BLACK SN850X NVMe™ SSD - 8TB, Without Heatsink inside. Below are my speed results:



I also like this enclosure because it comes with a silicon frame case that allows me to throw it in a backpack with my 16" MacBookPro without further protection.

 

[Jacura9]

Using a Mac Studio M3 Ultra (standard config with 28 core CPU, 96 GB RAM and 1 TB HD), additional storage was required. The solution was to employ an external M.2 drive in a Thunderbolt 5 enclosure for maximum speed.

There are few choices in external drives with TB5 connectivity. A couple of complete drives and a couple of enclosures were all that were available for immediate order.

Of those, an enclosure from Acasis was the only one incorporating a cooling fan so it was procured, model number TB501Pro, advertising 80 Gb/s speed.

The requirements for the M.2 NVME drive are specific. It must be a 4 lane stick employing PCIe Gen 4. The enclosure does not support Gen 5.

On hand was a WD_Black SN850X NVMe™ SSD, 2TB. It fit all the requirements and seemed fast enough so it was installed in the Acasis.

The drive was plugged into the Mac using the Acasis supplied cable and formatted with the Apple disk utility in APFS to match the internal drive. Approximately a half TB of miscellaneous data files were then transferred to the drive prior to testing to better simulate real world conditions.

Blackmagic’s Disk Speed Test was used to check the throughput. This software was selected due to familiarity.

The results are noteworthy. To begin with, both the internal and external drives wrote faster than they read. No fluke, repeated testing and reinstalling the testing software yielded exactly the same results.

Those results were fast. First, the internal 1 TB drive:


Internal Drive

Then the external 2 TB Thunderbolt 5 drive:


External Drive

Note: When testing, no two results were identical. The widest variation was on the order of 8%. The results shown are “typical” in that they are close to the mean.

To complete the testing a 138.1 GB .NEV file was transferred from the external to the internal drive using a Finder copy. Timing the transfer was considered accurate to within 0.1 second.

The transfer took 22.93 seconds which results in a 6.023 GB/S transfer speed. This is actually slightly faster than the results from the Blackmagic testing (taking the lower read speed into account), likely due to the single monolithic file transfer.

Transferring the same file from the internal to the external drive took slightly longer, resulting in a transfer speed of 5.4 GB/S. Speculation was that this was due to the slower read speed of the internal drive, however this is just an uninformed guess.

While the TB5 enclosures are still quite expensive (the Acasis cost ~ $225 U.S., plus the cost of the drive) it is still far less costly than Apple’s drive pricing.

With the TB5 drive closely matching the speed of the internal drive it makes sense to go this route rather than paying Apple. NB: The larger drives provided by Apple may or may not be significantly faster than the 1 TB drive contained in the test Mac Studio. Any info on this would be appreciated.

In comparison a PC equipped with a USB-C 3.2 Gen 2x2 card and an appropriate drive and cable attached delivered 1.8 GB/S under similar test conditions. This amounts to 72% of the theoretical 20 Gb/S speed available.

TB5 appears to be three times faster while using only 60% of the theoretical throughput. Perhaps efficiency gains in the future will provide even more speed.

Certainly there is much to criticize in the testing methods employed. The write speeds that are faster than the read speeds is unusual. However experience in real world usage shows that the tested speeds are realized in practice. Perceived speeds of the internal and external drives are essentially identical.

Input from anyone with TB5 drive experience is welcome.

That's certainly not slow. It seems the system bus is the limiting factor now. We've come a long way from the SATA spinning drives.

 

[Jacura9]

It seems the $200 price point is pretty normal for this class. I just need the 8TB NVMe to drop in price. The 4TB are ubiquitous and cheap but 8TB are a big jump still.

 

[Mathew Anderson]

Using a Mac Studio M3 Ultra (standard config with 28 core CPU, 96 GB RAM and 1 TB HD), additional storage was required. The solution was to employ an external M.2 drive in a Thunderbolt 5 enclosure for maximum speed.

There are few choices in external drives with TB5 connectivity. A couple of complete drives and a couple of enclosures were all that were available for immediate order.

Of those, an enclosure from Acasis was the only one incorporating a cooling fan so it was procured, model number TB501Pro, advertising 80 Gb/s speed.

The requirements for the M.2 NVME drive are specific. It must be a 4 lane stick employing PCIe Gen 4. The enclosure does not support Gen 5.

On hand was a WD_Black SN850X NVMe™ SSD, 2TB. It fit all the requirements and seemed fast enough so it was installed in the Acasis.

The drive was plugged into the Mac using the Acasis supplied cable and formatted with the Apple disk utility in APFS to match the internal drive. Approximately a half TB of miscellaneous data files were then transferred to the drive prior to testing to better simulate real world conditions.

Blackmagic’s Disk Speed Test was used to check the throughput. This software was selected due to familiarity.

The results are noteworthy. To begin with, both the internal and external drives wrote faster than they read. No fluke, repeated testing and reinstalling the testing software yielded exactly the same results.

Those results were fast. First, the internal 1 TB drive:


Internal Drive

Then the external 2 TB Thunderbolt 5 drive:


External Drive

Note: When testing, no two results were identical. The widest variation was on the order of 8%. The results shown are “typical” in that they are close to the mean.

To complete the testing a 138.1 GB .NEV file was transferred from the external to the internal drive using a Finder copy. Timing the transfer was considered accurate to within 0.1 second.

The transfer took 22.93 seconds which results in a 6.023 GB/S transfer speed. This is actually slightly faster than the results from the Blackmagic testing (taking the lower read speed into account), likely due to the single monolithic file transfer.

Transferring the same file from the internal to the external drive took slightly longer, resulting in a transfer speed of 5.4 GB/S. Speculation was that this was due to the slower read speed of the internal drive, however this is just an uninformed guess.

While the TB5 enclosures are still quite expensive (the Acasis cost ~ $225 U.S., plus the cost of the drive) it is still far less costly than Apple’s drive pricing.

With the TB5 drive closely matching the speed of the internal drive it makes sense to go this route rather than paying Apple. NB: The larger drives provided by Apple may or may not be significantly faster than the 1 TB drive contained in the test Mac Studio. Any info on this would be appreciated.

In comparison a PC equipped with a USB-C 3.2 Gen 2x2 card and an appropriate drive and cable attached delivered 1.8 GB/S under similar test conditions. This amounts to 72% of the theoretical 20 Gb/S speed available.

TB5 appears to be three times faster while using only 60% of the theoretical throughput. Perhaps efficiency gains in the future will provide even more speed.

Certainly there is much to criticize in the testing methods employed. The write speeds that are faster than the read speeds is unusual. However experience in real world usage shows that the tested speeds are realized in practice. Perceived speeds of the internal and external drives are essentially identical.

Input from anyone with TB5 drive experience is welcome.

I have that same drive, it's fantastic. Internal connection in this case.

 

[LarryRC]

Using a Mac Studio M3 Ultra (standard config with 28 core CPU, 96 GB RAM and 1 TB HD), additional storage was required. The solution was to employ an external M.2 drive in a Thunderbolt 5 enclosure for maximum speed.

There are few choices in external drives with TB5 connectivity. A couple of complete drives and a couple of enclosures were all that were available for immediate order.

Of those, an enclosure from Acasis was the only one incorporating a cooling fan so it was procured, model number TB501Pro, advertising 80 Gb/s speed.

The requirements for the M.2 NVME drive are specific. It must be a 4 lane stick employing PCIe Gen 4. The enclosure does not support Gen 5.

On hand was a WD_Black SN850X NVMe™ SSD, 2TB. It fit all the requirements and seemed fast enough so it was installed in the Acasis.

The drive was plugged into the Mac using the Acasis supplied cable and formatted with the Apple disk utility in APFS to match the internal drive. Approximately a half TB of miscellaneous data files were then transferred to the drive prior to testing to better simulate real world conditions.

Blackmagic’s Disk Speed Test was used to check the throughput. This software was selected due to familiarity.

The results are noteworthy. To begin with, both the internal and external drives wrote faster than they read. No fluke, repeated testing and reinstalling the testing software yielded exactly the same results.

Those results were fast. First, the internal 1 TB drive:


Internal Drive

Then the external 2 TB Thunderbolt 5 drive:


External Drive

Note: When testing, no two results were identical. The widest variation was on the order of 8%. The results shown are “typical” in that they are close to the mean.

To complete the testing a 138.1 GB .NEV file was transferred from the external to the internal drive using a Finder copy. Timing the transfer was considered accurate to within 0.1 second.

The transfer took 22.93 seconds which results in a 6.023 GB/S transfer speed. This is actually slightly faster than the results from the Blackmagic testing (taking the lower read speed into account), likely due to the single monolithic file transfer.

Transferring the same file from the internal to the external drive took slightly longer, resulting in a transfer speed of 5.4 GB/S. Speculation was that this was due to the slower read speed of the internal drive, however this is just an uninformed guess.

While the TB5 enclosures are still quite expensive (the Acasis cost ~ $225 U.S., plus the cost of the drive) it is still far less costly than Apple’s drive pricing.

With the TB5 drive closely matching the speed of the internal drive it makes sense to go this route rather than paying Apple. NB: The larger drives provided by Apple may or may not be significantly faster than the 1 TB drive contained in the test Mac Studio. Any info on this would be appreciated.

In comparison a PC equipped with a USB-C 3.2 Gen 2x2 card and an appropriate drive and cable attached delivered 1.8 GB/S under similar test conditions. This amounts to 72% of the theoretical 20 Gb/S speed available.

TB5 appears to be three times faster while using only 60% of the theoretical throughput. Perhaps efficiency gains in the future will provide even more speed.

Certainly there is much to criticize in the testing methods employed. The write speeds that are faster than the read speeds is unusual. However experience in real world usage shows that the tested speeds are realized in practice. Perceived speeds of the internal and external drives are essentially identical.

Input from anyone with TB5 drive experience is welcome.

That's certainly not slow. It seems the system bus is the limiting factor now. We've come a long way from the SATA spinning drives.

Actually the system bus speed of the M4 macbook pro starts at 273GB/s to 546 GB/s depending on the exact configuration. Meanwhile thunderbolt 5 is approximately 8GB/s to 12GB/s. So the limiting factor will be Thunderbolt 5.